A DeepSeek Wake-Up Call?
The market’s reaction to China’s DeepSeek AI model demonstrates that technology innovations can occur at any time, and unexpectedly, and those surprises can significantly disrupt stock prices for competitors. According to reports, its AI models use fewer advanced chips and this puts into question the extent to which growth will continue for US suppliers of those chips (Nvidia, Broadcom, Marvell, and others). In addition, if it is the case that DeepSeek is using far fewer chips, the world may need a smaller data center buildout and those data centers could need less energy. Yesterday, shares also fell in companies supplying equipment and energy to data centers. Investors seem now to be questioning risky assets more generally as hopes for growth in the leading sectors potentially become more muted. Does the arrival of DeepSeek call for an overhaul of portfolios? We don’t think so.
Valuations of the chip and data center infrastructure companies embody significant growth for many years to come, so any decrease in growth forecasts (or increase in the uncertainty around such forecasts) will reduce multiples investors are willing to pay today. It is too soon to determine the veracity of the claims being made by DeepSeek regarding its underlying hardware usage, and as new information arises in coming days and weeks we may very well see additional volatility in technology, data center infrastructure, and energy stocks.
But even if the current details of DeepSeek’s software prove true and become more widely adopted, imagine how much more powerful AI models would be using the vastly better hardware of the leading providers, including OpenAI, Microsoft, Amazon, and Google. Would they use this new approach to achieve even faster breakthroughs in their own AI models? It is possible, and perhaps likely, that competition among them would continue to push data center spending at its current pace, because which of them would want to be left behind? This remains the biggest unknown, and worried investors the most yesterday.
While uncertainty will likely surround growth outlooks related to chips and data centers stocks, advances in the software for training and applying AI models would lead to faster and more profitable adoption of AI applications. All this would bring clear benefits to firms and consumers who make use of AI in the future. And if AI-related services and products become less expensive, consumers would have more discretionary income to spread around to other purchases. This bodes well, not only for the software firms delivering those applications, but also for the firms using them, and for many consumer discretionary companies. So, ultimately, cheaper, better, faster AI is likely to sustain or even increase the pace of economic growth in the US (and beyond). That all suggests that companies tied to innovation and growth, more broadly, would remain an attractive space for investors over the long term.
Yesterday’s reaction also points to the value of diversifying: concentrated portfolios create unnecessary risks. And for the more risk-averse investor, or those with a shorter time horizon, the DeepSeek sell-off emphasizes the importance of layering into portfolios stocks offering higher dividend yields, cyclical resilience, and lower correlations with those in the technology sector.